KR20220085827A - Prediction systems, information processing devices and recording media - Google Patents

Abstract

The predictive model generating unit of the predictive system includes: means for determining one or a plurality of state values as explanatory variables from among a plurality of state values related to the learning sample used for generating the predictive model, based on the importance to the learning sample; Means for determining a section to be used for prediction by sequentially varying the sections included in the section and evaluating the prediction precision by the determined explanatory variable, and under the conditions of the determined explanatory variable and the determined section, the model parameters defining the predictive model and means for determining a model parameter for defining the predictive model by evaluating a plurality of indices with respect to the predictive model defined by each model parameter, sequentially and differently.

Description

Prediction systems, information processing devices and recording media

The present invention relates to a prediction system for predicting a change occurring in a control target, an information processing apparatus constituting the prediction system, and a recording medium storing an information processing program for realizing the information processing apparatus.

In various production sites, for some reason, an unusual change or an unusual change may occur. If the occurrence of such a change can be predicted in advance and some countermeasures can be taken, it is beneficial for maintaining the performance of production facilities, securing product quality, and the like.

Regarding such prior prediction, for example, Japanese Patent Application Laid-Open No. 2009-237832 (Patent Document 1) discloses a method for constructing a variable prediction model capable of improving the precision of demand prediction in all periods and seasons. The variable predictive model construction method disclosed in Patent Document 1 uses learning data obtained by adding correction to accumulated time series data to construct an appropriate predictive model for each of a plurality of learning periods of 7 to 70 days, and the modeling precision of each learning period By performing evaluation, the process of selecting the optimal learning period and prediction model with the highest prediction precision is employ|adopted.

Patent Document 1: Japanese Patent Laid-Open No. 2009-237832

However, in the variable predictive model construction method disclosed in Patent Document 1 described above, it is necessary to construct an appropriate predictive model for each of a plurality of learning periods, and then evaluate the modeling precision of each learning period, so that the optimal learning period and the predictive model There is a problem that the number of steps required to make a selection is large.

An object of the present invention is to provide a method capable of generating a predictive model more efficiently.

A prediction system according to an example of the present invention includes a control calculation unit that executes a control operation for controlling a control target, and inputting performance values including one or a plurality of state values among the state values that the control calculation unit can refer to into a prediction model. It includes a predictive value obtaining unit to be acquired, and a predictive model generating unit for determining a predictive model in advance. The predictive model generating unit includes means for determining one or a plurality of state values as explanatory variables from among the plurality of state values associated with the learning sample used for generating the predictive model, based on the importance to the learning sample, and included in the search section Means for determining an interval used for prediction by sequentially different intervals to be used for prediction by evaluating prediction precision by the determined explanatory variable, and under the conditions of the determined explanatory variable and the determined interval, the model parameters defining the predictive model are sequentially different and means for determining a model parameter for defining the predictive model by evaluating a plurality of indices with respect to the predictive model defined by each model parameter.

According to this structure, the predictive model which evaluated a some parameter|index can be produced|generated easily.

The predictive model generation unit may further include means for selecting a sample to be used as a learning sample from among a plurality of samples by clustering using one or a plurality of feature quantities calculated from each sample. According to this configuration, it is possible to prevent a plurality of samples similar to each other from being selected as samples for learning.

The predictive model generation unit may calculate the importance of each state value by applying a decision tree algorithm to the learning sample and a plurality of related state values. According to this configuration, by applying the decision tree algorithm, it is possible to easily determine which state value is suitable for the prediction model.

The predictive model generating unit may evaluate the performance of the predictive model generated from each pattern in which at least one of the length and position of the section included in the search section is different. According to this configuration, it is possible to easily find a suitable section included in the search section.

The predictive model generation unit may search for a section used for prediction after fixing the model parameters of the predictive model. According to this configuration, by temporarily fixing the model parameters of the predictive model, it is possible to reduce the amount of processing related to the search.

The predictive model generation unit may further include means for providing a first user interface that accepts a user operation for changing a section used for prediction. According to this configuration, the user can fine-tune the determined section while confirming, and the like.

The plurality of indexes may include at least one of prediction accuracy, model size, and processing speed. According to this configuration, an optimal predictive model can be generated in consideration of the operation by including an important factor in the index in the operation.

The predictive model generating unit may further include means for providing a second user interface for displaying a plurality of indices. According to this structure, a user can grasp|ascertain at a glance a plurality of determined indexes.

According to another example of the present invention, an information processing apparatus connected to a control apparatus is provided. The control device includes a control arithmetic unit that executes a control operation for controlling a control target, and a predicted value acquisition unit that acquires a predicted value by inputting an performance value comprising one or more state values among the state values that the control arithmetic unit can refer to into a predictive model. . The information processing device is a predictive model generator that determines a predictive model in advance, and among a plurality of state values associated with a learning sample used for generating the predictive model, based on the importance to the learning sample, one or a plurality of state values A means for determining as an explanatory variable; a means for determining a section to be used for prediction by sequentially differentiating the sections included in the search section and evaluating the prediction precision by the determined explanatory variable; , means for determining a model parameter for defining the predictive model by sequentially varying the model parameters defining the predictive model, and evaluating a plurality of indices with respect to the predictive model defined by each model parameter.

According to another example of the present invention, there is provided an information processing program that is executed on a computer connected to a control device. The control device includes a control arithmetic unit that executes a control operation for controlling a control target, and a predicted value acquisition unit that acquires a predicted value by inputting an performance value comprising one or more state values among the state values that the control arithmetic unit can refer to into a predictive model. . The information processing program is a process for pre-determining a predictive model, and, among a plurality of state values associated with the learning sample used for generation of the predictive model, in the computer, based on the importance to the learning sample, one or a plurality of states a step of determining the value as an explanatory variable, determining a section to be used for prediction by sequentially differentiating the sections included in the search section and evaluating the prediction precision by the determined explanatory variable, and the determined explanatory variable and the conditions of the determined section In the following, a step of determining a model parameter for defining a predictive model is executed by sequentially differentiating the model parameters defining the predictive model, and evaluating a plurality of indices with respect to the predictive model defined by each model parameter.

According to the present invention, a predictive model can be generated more efficiently.

BRIEF DESCRIPTION OF THE DRAWINGS It is a schematic diagram which shows the overall structural example of the prediction system which concerns on this embodiment.
2 is a schematic diagram showing an application example of the prediction system according to the present embodiment.
3 is a schematic diagram showing an example of control based on a prediction result by the prediction system according to the present embodiment.
4 is a flowchart showing a processing procedure of a prediction model generation processing using the prediction system according to the present embodiment.
Fig. 5 is a block diagram showing a hardware configuration example of a control device constituting the prediction system according to the present embodiment.
Fig. 6 is a block diagram showing a hardware configuration example of a support device constituting the prediction system according to the present embodiment.
Fig. 7 is a block diagram showing an example of a software configuration of a control device and a support device constituting the prediction system according to the present embodiment.
Fig. 8 is a block diagram showing an outline of a function module included in the analysis program shown in Fig. 7;
It is a figure for demonstrating the selection example of the sample for learning in the prediction system which concerns on this embodiment.
Fig. 10 is a flowchart showing a more detailed processing procedure regarding selection of a learning sample (step S3) in the processing procedure of the generation processing shown in Fig. 4;
Fig. 11 is a diagram showing an example of clustering regarding selection of samples for learning in the processing sequence of the generation processing shown in Fig. 4;
Fig. 12 is a diagram showing an example of a user interface provided in a process for selecting a sample for learning in the prediction system according to the embodiment;
Fig. 13 is a diagram showing another example of a user interface provided in a process for selecting a sample for learning in the prediction system according to the embodiment;
Fig. 14 is a flowchart showing a more detailed processing procedure regarding the determination of explanatory variables and sections (step S4) in the processing procedure of the generation processing shown in Fig. 4;
Fig. 15 is a schematic diagram for schematically illustrating the processing contents of the determination of explanatory variables and sections (step S4) in the processing procedure of the generation processing shown in Fig. 4;
Fig. 16 is a diagram showing an example of a user interface provided in processing for determining explanatory variables and sections in the prediction system according to the embodiment;
Fig. 17 is a diagram showing an example of a user interface provided in processing for determining explanatory variables and sections in the prediction system according to the embodiment;
Fig. 18 is a diagram showing an example of a user interface provided in processing for determining explanatory variables and sections in the prediction system according to the embodiment;
Fig. 19 is a diagram showing an example of a user interface provided in processing for determining explanatory variables and sections in the prediction system according to the embodiment;
Fig. 20 is a flowchart showing a more detailed processing procedure related to the determination of the model parameter (step S5) shown in Fig. 4;
Fig. 21 is a diagram showing an example of a user interface provided in a process for determining a model parameter in the prediction system 1 according to the embodiment;
22 is a schematic diagram showing an outline configuration of an abnormality detection system according to a modification of the present embodiment.
Fig. 23 is a diagram for explaining a sample selection processing example shown in Fig. 22;

EMBODIMENT OF THE INVENTION Embodiment of this invention is described in detail, referring drawings. In addition, about the same or equivalent part in a figure, the same code|symbol is attached|subjected and the description is not repeated.

<A. Application example>

First, an example of a scene to which the present invention is applied will be described.

The main aspects of the control system with a prediction function according to the present embodiment will be described. In the following description, the entire control system is also referred to as a "prediction system" because the explanation is mainly focused on the prediction function that the control system has.

1 : is a schematic diagram which shows the overall structural example of the prediction system 1 which concerns on this embodiment. Referring to FIG. 1 , the prediction system 1 according to the present embodiment includes, as main components, a control device 100 for controlling a control target, and a support device 200 connected to the control device 100 . include

The control device 100 may be embodied as a kind of computer such as a programmable controller (PLC). The control device 100 may be connected to the field device group 10 via the field bus 2 and also to one or more display devices 400 via the field bus 4 . In addition, the control device 100 may be connected to the upper level server 300 via the upper level network 6 . In addition, the host server 300 and the display device 400 are optional components, and are not essential components of the prediction system 1 .

The control device 100 has a control logic (hereinafter also referred to as a "PLC engine") that executes various calculations for controlling equipment and machines. In addition to the PLC engine, the control device 100 has a collection function that collects data measured by the field device group 10 and transmitted to the control device 100 (hereinafter also referred to as “input data”). . Moreover, the control apparatus 100 also has a prediction function which predicts future time change based on the collected input data.

Specifically, a time series database (hereinafter, also referred to as "Time Series Data Base (TSDB)") 130 mounted on the control device 100 provides a collection function, and a prediction mounted on the control device 100 . Model 140 provides the monitoring function. Details of the TSDB 130 and the prediction model 140 will be described later.

It is preferable that the field bus 2 and the field bus 4 employ|adopt an industrial communication protocol. As such a communication protocol, EtherCAT (registered trademark), EtherNet/IP (registered trademark), DeviceNet (registered trademark), CompoNet (registered trademark), and the like are known.

The field device group 10 includes devices that collect input data from a control target or control-related manufacturing device, production line, or the like (hereinafter also referred to as "field"). As a device for collecting such input data, an input relay, various sensors, and the like are assumed. The field device group 10 further includes devices that give a certain action to the field based on a command generated by the control device 100 (hereinafter also referred to as “output data”). An output relay, a contactor, a servo driver, a servo motor, and any actuator other than an output relay, a contactor, are assumed as a device which gives a certain action to such a field. These field device groups 10 exchange data including input data and output data with the control device 100 via the field bus 2 .

In the configuration example shown in FIG. 1 , the field device group 10 includes a remote I/O (Input/Output) device 12 , a relay group 14 , an image sensor 18 , and a camera 20 , , including a servo driver 22 and a servo motor 24 .

The remote I/O device 12 includes a communication unit for communicating via the field bus 2 and an input/output unit for collecting input data and outputting output data (hereinafter, also referred to as “I/O unit”). ) is included. Through such an I/O unit, input data and output data are exchanged between the control device 100 and the field. 1 shows an example in which digital signals are exchanged as input data and output data via the relay group 14 .

The I/O unit may be directly connected to the field bus. 1 shows an example in which the I/O unit 16 is directly connected to the field bus 2 .

The image sensor 18 performs image measurement processing such as pattern matching on the image data captured by the camera 20 , and transmits the processing result to the control device 100 .

The servo driver 22 drives the servo motor 24 according to output data (eg, a position command, etc.) from the control device 100 .

As described above, data is exchanged between the control device 100 and the field device group 10 via the field bus 2, and these data exchanged are extremely small in the order of several hundred μsec to several tens of msec. It is updated in a short period. In addition, such an update process of data exchanged is sometimes referred to as "I/O refresh process".

In addition, the display device 400 connected to the control device 100 via the field bus 4 receives an operation from the user and transmits a command or the like in response to the user operation to the control device 100 , The calculation result and the like in the control device 100 are graphically displayed.

The upper server 300 is connected to the control device 100 through the upper network 6 , and exchanges necessary data with the control device 100 . In the upper network 6, a general-purpose protocol such as Ethernet (registered trademark) may be implemented.

The support apparatus 200 is an information processing apparatus (an example of a computer) that supports the preparation necessary for the control apparatus 100 to control the control object. Specifically, the support device 200 is connected to the development environment (program creation and editing tool, parser, compiler, etc.) of the user program executed in the control device 100 , the control device 100 , and the control device 100 . A setting environment for setting parameters (configuration) of various devices, a function to transmit a generated user program to the control device 100, a function to modify/change a user program executed on the control device 100 online, etc. , etc. are provided.

In addition, the support device 200 according to the present embodiment has a function for supporting the generation and optimization of the predictive model 140 to be implemented in the control device 100 . That is, the support apparatus 200 has a predictive model generator that determines the predictive model 140 in advance. Details of these functions will be described later.

Next, an application example of the control device 100 included in the prediction system 1 will be described.

2 is a schematic diagram showing an application example of the prediction system 1 according to the present embodiment. 2 shows an example of a production facility including a press machine 30 .

Referring to FIG. 2 , the press machine 30 receives the work 31 and arranges the received work 31 on the support 34 provided on the base 33 . Then, the work 31 is compressed with a press-fit plate 35 provided at the tip of the drive shaft 36 driven by the motor 37 to produce the intermediate product 32 .

In the press machine 30, it is said that defects may occur in the intermediate product 32 due to unexpected factor fluctuations. For this reason, it is determined whether defects have occurred in the intermediate product 32 by inspection by an inspection machine arranged on the downstream side of the press machine 30, or by visual inspection by an inspector or inspection by extraction. judge If it is determined that a defect is occurring, the target value or the like is adjusted each time.

As described above, in a normal manufacturing process, in order to maintain and improve the yield rate of the intermediate product 32, it is necessary to adjust the target value each time.

On the other hand, by predicting the state of the intermediate product 32 (that is, the quality after processing) using the predictive model 140 according to the present embodiment, before a defect actually occurs, the Control can be calibrated. By being able to utilize such a prediction of occurrence of a defect in advance, it is possible to reduce the man-hours for each time adjustment of a target value and the like, and to prevent a defect from occurring in the intermediate product 32 .

3 is a schematic diagram illustrating an example of control based on a prediction result by the prediction system 1 according to the present embodiment.

3(A) shows the planned value (command) of the press-in position of the press machine at a certain point in time, and the actual press-in position (actual value) of the press machine 30 . The target value represents the desired thickness of the intermediate product 32 after processing.

Referring to Fig. 3(B), at a certain point in time, based on the information (which may include actual values) up to that point, the press-fitting position (predicted value) of the press machine 30 in the future is calculated, and the calculated predicted value is Based on it, you may make it correct|amend the control amount with respect to the press machine 30. As shown in FIG.

In the press machine 30 shown in FIG. 2 and FIG. 3, the press-fit position of the press machine 30 is predicted, and the control amount is corrected based on the prediction result, for example, in response to the dispersion|variation in hardness of the workpiece|work 31, etc. , it is not necessary for the operator to adjust the target value each time. As a result, generation of defective products due to unexpected factor fluctuations can be suppressed, and even if there is any variation in the workpiece 31, quality can be stabilized.

Data used for prediction (actual values or observation values) and predicted data may be partially or entirely the same, or may be completely different from each other.

The prediction system 1 according to the present embodiment provides a function for appropriately generating the prediction model 140 . A function for appropriately generating the predictive model 140 may be typically implemented in the support apparatus 200 .

<B. Overview of the creation and operation of predictive models>

Next, the outline|summary of generation|occurrence|production and operation of the prediction model 140 using the prediction system 1 which concerns on this embodiment is demonstrated.

4 is a flowchart showing a processing procedure of the generation processing of the prediction model 140 using the prediction system 1 according to the present embodiment. Each step shown in FIG. 4 is typically realized by the processor 202 of the support device 200 executing a program (analysis program 226, OS 228, etc.).

When FIG. 4 is referred, the support apparatus 200 acquires the time series data of the performance value stored in TSDB 130 (step S1). Next, the support apparatus 200 accepts the setting of the prediction target section from the time series data of the acquired performance value (step S2).

The support apparatus 200 selects the learning sample used for generation|generation of the prediction model for predicting the change of the prediction object section set in step S2 (step S3). In step S3, from among the plurality of types of data, which data is to be used for learning is selected. More specifically, the support apparatus 200 determines a class by clustering for a statistic (feature amount) of each time series data (waveform sample), and extracts target data from the determined class. The processing of step S3 will be described in detail.

Next, the support apparatus 200 determines an explanatory variable and a section (step S4). And the support apparatus 200 determines the model parameter to be used (step S5). In step S5, the support apparatus 200 generates the predictive model 140 according to the explanatory variable and the section selected in step S4, and evaluates the performance of the generated predictive model 140 to select an appropriate predictive algorithm. Explore. The support apparatus 200 determines as a model parameter using a prediction algorithm with high performance of prediction precision and prediction speed.

Finally, the support apparatus 200 generates the predictive model 140 based on the determined model parameters and the like (step S6).

By setting the predictive model 140 generated by the above processing procedure in the control device 100, the operation shown in Figs. 2 and 3 is enabled.

<C. Hardware configuration example>

Next, a hardware configuration example of a main device constituting the prediction system 1 according to the present embodiment will be described.

(c1: hardware configuration example of control device 100)

5 is a block diagram showing an example of the hardware configuration of the control device 100 constituting the prediction system 1 according to the present embodiment. Referring to FIG. 5 , the control device 100 includes a processor 102 such as a central processing unit (CPU) or a micro-processing unit (MPU), a chip set 104 , a main memory device 106 , and 2 The car storage device 108 , the upper network controller 110 , the USB (Universal Serial Bus) controller 112 , the memory card interface 114 , the internal bus controller 122 , the field bus controller 118 , 120) and I/O units 124-1, 124-2, ....

The processor 102 reads various programs stored in the secondary storage device 108, expands them to the main storage device 106 and executes them, thereby generating the PLC engine 150 (refer to FIG. 7) and the predictive model 140. come true The chip set 104 controls data transfer and the like between the processor 102 and each component.

In the secondary storage device 108 , in addition to the system program for realizing the PLC engine 150 , a user program executed using the PLC engine 150 is stored. The secondary storage device 108 also stores a program for realizing the predictive model 140 .

The upper network controller 110 controls data exchange with other devices through the upper network 6 . The USB controller 112 controls data exchange with the support device 200 via a USB connection.

The memory card interface 114 is configured such that the memory card 116 is detachable, writes data to and from the memory card 116, and receives various data (such as user programs and trace data) from the memory card 116 . It is possible to read.

The internal bus controller 122 is an interface for exchanging data with the I/O units 124-1, 124-2, ... mounted on the control device 100.

The field bus controller 118 controls data exchange with other devices via the field bus 2 . Similarly, the field bus controller 120 controls data exchange with other devices via the field bus 4 .

5, the processor 102 shows a configuration example in which necessary functions are provided by executing a program, some or all of these provided functions are provided by a dedicated hardware circuit (eg, ASIC (Application Specific Integrated Circuit)) Alternatively, it may be mounted using an FPGA (Field-Programmable Gate Array, etc.). Alternatively, the main part of the control device 100 may be realized using hardware according to a general-purpose architecture (eg, an industrial personal computer based on a general-purpose personal computer). In this case, a plurality of operating systems (OSs) having different uses may be executed in parallel by using virtualization technology, and a necessary application may be executed on each OS.

(c2: hardware configuration example of support device 200)

The support device 200 according to the present embodiment is realized by executing a program using, as an example, hardware according to a general-purpose architecture (eg, a general-purpose personal computer).

6 is a block diagram showing an example of the hardware configuration of the support apparatus 200 constituting the prediction system 1 according to the present embodiment. Referring to FIG. 6 , the support device 200 includes a processor 202 such as a CPU or MPU, an optical drive 204 , a main memory device 206 , a secondary storage device 208 , and a USB controller ( 212 , a higher-level network controller 214 , an input unit 216 , and a display unit 218 . These components are connected via a bus 220 .

The processor 202 reads various programs stored in the secondary storage device 208, expands them to the main storage device 206 and executes them, thereby realizing various processes including a model generation process to be described later.

The secondary storage device 208 is constituted of, for example, a hard disk drive (HDD), a flash solid state drive (SSD), or the like. The secondary storage device 208 typically includes a development program 222 for creating a user program to be executed on the support device 200, debugging the created user program, defining the system configuration, setting various parameters, and the like. ) and the PLC interface program 224 for exchanging data related to the prediction function between the control device 100, the analysis program 226 for realizing generation of the prediction model 140, etc., and the OS 228 ) is stored. In the secondary storage device 208, a necessary program other than the program shown in FIG. 6 may be stored.

The support device 200 has an optical drive 204, and a recording medium 205 (for example, a DVD (Digital Versatile Disc), etc.) which non-transiently stores a computer readable program. of the optical recording medium), the program stored therein is read and installed in the secondary storage device 208 or the like.

The various programs executed by the support device 200 may be installed via the computer-readable recording medium 205, but may be installed in the form of being downloaded from an arbitrary server on a network. In addition, the function provided by the support apparatus 200 which concerns on this embodiment may be implemented in the form using a part of module provided by OS.

The USB controller 212 controls data exchange with the control device 100 through a USB connection. The upper network controller 214 controls data exchange with other devices through an arbitrary network.

The input unit 216 is constituted by a keyboard, a mouse, or the like, and accepts user operations. The display unit 218 includes a display, various indicators, a printer, and the like, and outputs processing results and the like from the processor 202 .

6 shows a configuration example in which necessary functions are provided when the processor 202 executes a program, some or all of these provided functions using a dedicated hardware circuit (eg, ASIC or FPGA) So you can mount it.

<D. Software configuration example/function configuration example>

Next, a software configuration example and a functional configuration example of the control device 100 and the support device 200 constituting the prediction system 1 according to the present embodiment will be described.

7 is a block diagram showing an example of the software configuration of the control device 100 and the support device 200 constituting the prediction system 1 according to the present embodiment.

Referring to FIG. 7 , the control device 100 includes a TSDB 130 and a predictive model 140 in addition to the PLC engine 150 as main functional components.

The PLC engine 150 sequentially interprets the user program 154 and executes the specified control operation. The PLC engine 150 manages the state value collected from the field in the form of a variable 152 , and the variable 152 is updated at a predetermined cycle. The PLC engine 150 may be realized when the processor 102 of the control device 100 executes a system program.

In this specification, the "state value" includes an input value collected from a field, a command value output to the field, and a system state value or internal value managed inside the control device 100 . In the control device 100 according to the present embodiment, “state value” is referred to in the form of “variable”. In the following description, for convenience, the term “variable” is used to include “state value”. use it In addition, the technical scope of this invention is not limited to the structure which refers to "state value" in the form of "variable".

The user program 154 includes a predicted value acquisition code 156 , an error evaluation code 158 , an additional learning code 160 , a TSDB recording code 162 , and a control operation code 164 .

The predicted value acquisition code 156 implements a predicted value acquisition unit that acquires a predicted value by inputting into the predictive model 140 an performance value composed of one or more state values among the state values that the control operation code 164 can refer to. More specifically, the predicted value acquisition code 156 includes an instruction for acquiring a predicted value by acquiring a necessary performance value managed as the variable 152 and inputting it to the predictive model 140 .

The error evaluation code 158 includes instructions for evaluating the error between the predicted value and the target value obtained by the predicted value acquisition code 156 .

The additional training code 160 includes instructions to further train the predictive model 140 , as needed, in response to the error assessed by the error evaluation code 158 . The additional training code 160 updates the model parameters 142 that define the predictive model 140 .

The TSDB recording code 162 acquires a predetermined variable among the variables managed as the variable 152 and records it in the storage area 132 of the TSDB 130 .

The control operation code 164 implements a control operation unit that executes a control operation for controlling a control object. More specifically, the control operation code 164 executes a control operation for controlling the control object, and in response to the error evaluated by the error evaluation code 158, if necessary, a target value used for the control operation. to correct

The TSDB 130 has an export module 134 that exports the data recorded in the storage area 132 to the support device 200 or the like as needed.

The predictive model 140 has a reference trajectory 144 in addition to the model parameters 142 that define the predictive model 140 .

On the other hand, in the support apparatus 200, the development program 222 and the analysis program 226 are installed.

The development program 222 generates the user program 154 according to a user operation, and transmits it to the control device 100 . In addition, the development program 222 also has a function of appropriately correcting the contents of the control operation code 164 .

The analysis program 226 corresponds to an information processing program for realizing the predictive model generation unit that determines the predictive model 140 in advance. More specifically, the analysis program 226 supports the generation of the predictive model 140, and includes an explanatory variable/interval determination module 2261, a model generation module 2262, an inference module 2263, evaluation module 2264 .

The explanatory variable/section determination module 2261 implements a function necessary for the process of determining the explanatory variable and the section (refer to step S4 shown in FIG. 4 ).

The model generation module 2262 implements a function necessary for the process of generating the predictive model 140 (refer to step S6 shown in FIG. 4 ).

The inference module 2263 performs inference (prediction) using the generated prediction model 140 and provides the prediction result to the evaluation module 2264 . The evaluation module 2264 evaluates the performance of the predictive model 140 of the object based on the prediction result from the inference module 2263 . By the functions provided by the inference module 2263 and the evaluation module 2264, a process for determining the model parameter to be used (refer to step S5 shown in Fig. 4) is realized.

FIG. 8 is a block diagram showing an outline of a function module included in the analysis program 226 shown in FIG. 7 . Referring to FIG. 8 , the analysis program 226 of the support device 200 includes a user interface 230 , an input/output management module 236 , a screen display module 238 , and a graph library as main functional configurations. 240 , an analysis module 242 , and an analysis library 244 .

The user interface 230 accepts settings from the user and performs general processing for providing various types of information to the user. As a specific implementation form, the user interface 230 has a script engine 232 , reads a setting file 234 containing a script describing a necessary process, and executes the set process.

The input/output management module 236 includes a file input function for reading data from a specified file or the like, a stream input function for receiving a data stream, and a file output function for outputting a file including generated data and the like.

The screen display module 238 includes a line graph generation function for generating a line graph based on input data or the like, and a parameter adjustment function for changing various parameters in response to a user's operation. Depending on the parameter change, the line graph generation function may update the line line. The line graph generation function and the parameter adjustment function refer to the graph library 240 and perform necessary processing.

The analysis module 242 is a module for realizing the main processing of the analysis program 226 and has a waveform sample function, an explanatory variable/section selection function, a parameter selection function, and a model generation function. Each function included in the analysis module 242 is realized by referring to the analysis library 244 .

The analysis library 244 includes a library for each function included in the analysis module 242 to execute processing. More specifically, the analysis library 244 includes a statistical function, a decision tree function, a time series regression function, a grid search function, a clustering function, an inference speed evaluation function, a precision evaluation function, and abnormality detection. It is good to have a function.

Hereinafter, the main processing for generating the predictive model 140 shown in FIG. 4 is demonstrated.

<E. Acquisition of time series data of performance value (step S1) and determination of prediction target section (step S2)>

In step S1 shown in FIG. 4, the support apparatus 200 acquires the time series data of the performance value stored in TSDB 130 of the control apparatus 100 according to user operation. The user sets the prediction target section while looking at the time series data of the performance value displayed on the support device 200 .

The prediction target section may be appropriately set in accordance with the characteristics of the control target shown in Figs. 2 and 3 described above.

<F. Selection of a learning sample used for generation of a predictive model (step S3)>

In step S3 shown in FIG. 4 , the support apparatus 200 selects a learning sample used for generation of the predictive model 140 .

In this specification, "sample" means a data string of a predetermined time length used as teacher data of prediction values to be output from the prediction model 140 . As for "sample", time series data (raw data) of a prediction target is basically used, but when a prediction target is a feature quantity extracted from time series data, you may use a characteristic quantity. The term "sample" pays attention to the processing unit at the time of processing a plurality of data, and the content of the data included therein is not particularly limited.

In this specification, data referred to for calculating or determining an arbitrary predictive value is also referred to as an "explanatory variable". An arbitrary predicted value is calculated or determined using one or a plurality of “explanatory variables”. Therefore, the learning sample is related in some way to data that can be a candidate for "explanatory variable".

In the present specification, "feature amount" is a term containing information contained in the time series data to be processed, for example, the maximum, minimum, median, average, standard deviation, variance, etc. of the time series data of the object. may include In addition, the "characteristic amount" may also include the target time series data itself.

9 : is a figure for demonstrating the selection example of the sample for learning in the prediction system 1 which concerns on this embodiment. Fig. 9 shows an example of time series data (sample) of a predetermined time length corresponding to the prediction target section.

Among these samples, the more different the change pattern is used for learning, the higher the prediction accuracy of the predictive model 140 is. That is, even if it learns using the sample which shows the same change pattern, it does not contribute to the improvement of the prediction precision of the predictive model 140. As shown in FIG.

As samples having different change patterns, in the example shown in Fig. 9, four samples surrounded by thick borders are selected as the learning samples.

As an example of a method of selecting samples with different change patterns, one or more feature quantities (eg, average values or standard deviations, etc.) are calculated from each sample, and based on the calculated one or more feature quantities, , cluster the samples. By such clustering, one or a plurality of classes included in the target sample group are extracted. Then, one or a plurality of samples belonging to each extracted class are respectively selected as a learning sample.

Fig. 10 is a flowchart showing a more detailed processing procedure regarding the selection of a learning sample (step S3) in the processing procedure of the generation processing shown in Fig. 4 .

Referring to FIG. 10 , the support apparatus 200 trims the target time series data to generate a sample (step S31). Trimming of time series data is performed by extracting only the portion related to the prediction target section.

Next, the support apparatus 200 calculates one or a plurality of feature quantities from each sample (step S32), and performs clustering based on the calculated one or more feature quantities (step S33).

The support apparatus 200 selects one or a plurality of samples from each of one or a plurality of classes determined by clustering, and determines the selected sample as a learning sample (step S34).

In this way, as a function related to the generation of the predictive model 140 , the support device 200 selects a sample used as a learning sample among a plurality of samples by clustering using one or a plurality of feature quantities calculated from each sample. Execute the selection process.

FIG. 11 is a diagram showing an example of clustering regarding selection of samples for learning in the processing sequence of the generation processing shown in FIG. 4 . Fig. 11 shows an example in which two characteristic quantities (average value and standard deviation) are calculated from each sample, and each sample is plotted in a two-dimensional space using each characteristic quantity as a coordinate.

It turns out that the sample group shown in FIG. 11 contains three clusters (class (1 - 3)) with respect to the characteristic amount of an average value and standard deviation. One or a plurality of samples belonging to each extracted class are respectively selected. That is, one or a plurality of samples belonging to class (1) are extracted, one or a plurality of samples belonging to class (2) are extracted, and one or a plurality of samples belonging to class (3) are extracted. Accordingly, a plurality of samples having different classes are determined as the learning samples.

The support apparatus 200 may automatically perform selection of the sample used as the target of a learning sample, and you may make a user support a process or confirm a selection result. For example, a user may confirm the sample selection result, and you may make it select again manually.

12 is a diagram showing an example of a user interface provided in a process for selecting a sample for learning in the prediction system 1 according to the embodiment. Fig. 13 is a diagram showing another example of a user interface provided in a process for selecting a sample for learning in the prediction system 1 according to the embodiment. 12 and 13 show an example of displaying the result of sample selection.

The user interface screen 250 shown in Fig. 12 can display the synthesized waveform 252 in which checked samples from the selected sample group 254 are overlapped on the same time axis and displayed. By selecting an arbitrary sample from the sample group 254 and displaying it as the synthesized waveform 252, the user can easily confirm whether the sample is selected or not.

The user interface screen 260 shown in FIG. 13 displays the degree of similarity between the selected samples in a matrix format. The user interface screen 260 has display elements arranged in a matrix form, and waveforms of each sample are arranged in one diagonal form. Above the diagonal line, the degree of similarity 264 (correlation coefficient) between samples is shown, and below the diagonal line, a synthesized waveform 262 in which two samples are displayed overlapping each other on the same time axis is displayed. .

The user interface screens shown in FIGS. 12 and 13 may be displayed on the same screen. In this case, when the user selects an arbitrary sample on the user interface screen 250, a corresponding sample on the user interface screen 260 shown in FIG. 13 may be highlighted and displayed in response to the selected sample. In addition, by selecting an arbitrary waveform on the user interface screen 250, the selection target and the non-selection target may be switched alternately.

By the above processing, the learning sample used for generation|generation of a predictive model is selected.

<G. Determination of explanatory variables and sections (step S4)>

In step S4 shown in FIG. 4 , the support apparatus 200 determines an explanatory variable and a section regarding the predictive model 140 .

In step S4, with respect to the prediction algorithm employed in the predictive model 140, the predictive model 140 is generated and performance is evaluated with respect to the combination in which the explanatory variable and the interval are changed respectively after being fixed to a default value or the like. By doing so, the explanatory variables and intervals to be used are determined. That is, the support apparatus 200 fixes the model parameters of the prediction model 140 , and then searches for explanatory variables and sections used for prediction.

Regarding the explanatory variable to be used and the search section for determining the section, the control cycle set in the user program executed by the control device 100 is referred to, and a preferred search section is determined based on the set control cycle. good night.

Fig. 14 is a flowchart showing a more detailed processing procedure regarding the determination of explanatory variables and sections (step S4) in the processing procedure of the generation processing shown in Fig. 4 . Referring to FIG. 14 , the support device 200 obtains a control cycle with reference to the user program being executed in the target control device 100 (step S41). When the support device 200 is connected to the control device 100 , it is sufficient to access the control device 100 to obtain necessary information, and when the support device 200 is not connected to the control device 100 . You may make it acquire a control period with reference to the project etc. which the support apparatus 200 hold|maintains.

The support device 200 sets an integer multiple of the acquired control period as a search section for searching a section (step S42).

The support apparatus 200 selects one or a plurality of explanatory variables from a plurality of variables related to the preselected learning sample (refer to above-mentioned step S3) (step S43). More specifically, the support apparatus 200 applies the decision tree algorithm to the learning sample and a plurality of related variables, and calculates the importance of each variable. And the support apparatus 200 selects the variable with high importance as an explanatory variable.

In addition, as the decision tree algorithm, any known algorithm can be employed, for example, a random forest or the like can be used.

In this way, as a function related to the generation of the predictive model 140 , the support device 200 includes a plurality of variables (state values) related to the learning sample used to generate the predictive model 140 for the learning sample. A process for determining one or a plurality of variables (state values) as explanatory variables based on the importance level is executed.

The support apparatus 200 sets any one candidate section included in the search section (step S44). The candidate section is generated by varying the length and/or position of sections included in the search section. And the support apparatus 200 produces|generates the predictive model 140 using the sample for learning, using one or a plurality of explanatory variables selected in step S43, and the candidate section set in step S44 as parameters (step S45). ). Then, the support apparatus 200 evaluates the performance related to the inference of the generated predictive model 140 (step S46). In step S46, a numerical value indicating performance is stored in association with a corresponding candidate section.

Then, the support apparatus 200 determines whether the generation and performance evaluation of the prediction model 140 for all candidate sections included in the search section are completed (step S47). If the generation and performance evaluation of the prediction model 140 for all candidate sections included in the search section are not completed (NO in step S47), the support device 200 sets another candidate section included in the search section, (Step S48), the processes following step S45 are repeated.

On the other hand, if the generation and performance evaluation of the prediction model 140 for all candidate sections included in the search section are completed (YES in step S47), the support device 200 predicts the candidate section showing the highest performance evaluation. It is determined as a section of the model 140 (step S49).

As such, the support apparatus 200, as a function related to the generation of the predictive model 140, sequentially varies the sections included in the search section, and evaluates the prediction precision by the determined explanatory variable, thereby selecting the section used for prediction. The decision processing is executed. In the process of determining the section used for this prediction, the support device 200 evaluates the performance of the prediction model generated from each pattern in which at least one of the length and position of the section included in the search section is different.

FIG. 15 is a schematic diagram for schematically illustrating the processing contents of the determination of explanatory variables and sections (step S4) in the processing sequence of the generation processing shown in FIG. 4 . 15 , a plurality of variables 1, 2, ... By applying the decision tree algorithm to , n, the explanatory variables are determined. In the application of the decision tree algorithm, a learning sample is used.

In addition, a search section is determined based on the user program, and one or a plurality of candidate sections are set in the determined search section. Then, the prediction model 140 is generated for each candidate section, and the performance of the generated prediction model 140 is evaluated.

Finally, explanatory variables and intervals corresponding to the high-performance predictive model 140 are determined.

In this way, in step S4 shown in FIG. 4 , the support apparatus 200 determines an explanatory variable based on the importance calculated by the decision tree algorithm, and grid-searches a section within the search section using the determined explanatory variable. By (full search), the explanatory variables and intervals are determined.

In addition, in the process sequence of the generation processing shown in Fig. 14 described above, in step S43, one or a plurality of explanatory variables are first determined, but some candidates are also set for the explanatory variables, and each candidate (each candidate is 1 or (including a plurality of explanatory variables) may be made to execute the processing following step S44.

16 to 19 are diagrams showing an example of a user interface provided in the processing for determining explanatory variables and sections in the prediction system 1 according to the embodiment.

16 to 19 , the user interface screen 270 includes a waveform display area 276 for displaying performance values and predicted values.

In the waveform display area 276, waveforms (time-series data) of actual values and predicted values are displayed. The performance value is a waveform (time-series data) of any one sample included in the learning sample. The predicted value is a waveform (time-series data) output by the generated prediction model 140 .

With respect to the predetermined prediction point 271 , the position and time width of the section 272 defining the range of the explanatory variable input to the prediction model 140 may be arbitrarily set. The user can arbitrarily change the section width 273 of the section 272 and the time difference 274 with the prediction point by operating a mouse or the like to move the position of the section 272 .

In response to the section 272 arbitrarily set by the user, a waveform (time-series data) of a predicted value is calculated, and the waveform (time-series data) of the calculated predicted value is displayed in the waveform display area 276 .

The user interface screen 270 includes a list of explanatory variables 277 , and a checked variable among one or a plurality of variables is used as the explanatory variable.

From the user interface screen 270 shown in Figs. 16 to 19, it can be seen that the setting of the user interface screen 270 shown in Fig. 18 calculates the predicted value with the highest precision.

The user interface screen 270 includes a prediction error 275 indicating a deviation between the performance value and the predicted value.

In response to the user operating the user interface screen 270, the support apparatus 200 may calculate a predicted value using the prediction model 140 generated by the process shown in FIG. 14 as it is. Alternatively, in response to the user operating the user interface screen 270 to appropriately change the explanatory variable and/or the section, the support device 200 performs processing after step S43 in FIG. 14 , or after step S44 You may make the process run again. That is, the predictive model 140 may be regenerated.

By regenerating the predictive model 140 in response to the user interface screen 270 shown in FIGS. 16 to 19 , the user can determine while searching for explanatory variables and/or sections to be set in the predictive model 140 . have.

In this way, the support apparatus 200 provides the user interface screen 270 (refer FIGS. 16-19) for accepting a user operation for changing the section used for prediction.

By the above processing, the explanatory variable and the interval used for generation of the predictive model are determined.

<H. Determination of model parameters (step S5)>

In step S5 shown in FIG. 4 , the support apparatus 200 determines the model parameter used for the predictive model 140 .

In step S4 described above, the explanatory variable and the section are determined by fixing the prediction algorithm or the like employed in the predictive model 140 to a default value or the like. In step S5, on the premise of the determined explanatory variable and section, the model parameters of the predictive model 140 are determined while considering operational indices such as prediction precision, model size, and processing speed.

In the following description, prediction accuracy, model size, and processing speed are described as examples of a plurality of indices, but they are not necessarily limited to these factors. However, it is preferable that at least one of a prediction precision, a model size, and a processing speed is included as a some parameter|index.

Fig. 20 is a flowchart showing a more detailed processing procedure related to the determination of the model parameter (step S5) shown in Fig. 4 . Referring to FIG. 20 , the support apparatus 200 determines the search range of the model parameter (step S51). The search range of the model parameter determines the search range (upper and lower limits) of each parameter in addition to the type of the parameter to be searched.

Next, the support apparatus 200 selects as a target the data set of arbitrary model parameters included in the search range determined in step S51 (step S52). Then, the support apparatus 200 generates the predictive model 140 based on the data set of the selected model parameter (step S53), and the performance of the generated predictive model 140 (prediction precision, model size, processing speed) etc.) is evaluated (step S54).

The support apparatus 200 determines whether there is a data set of another model parameter included in the search range determined in step S51 (step S55). If there is a data set of another model parameter included in the search range determined in step S51 (YES in step S55), the support apparatus 200 selects a data set of another model parameter included in the search range determined in step S51 (step S56), and the processes following step S53 are repeated.

On the other hand, if there is no data set of other model parameters included in the search range determined in step S51 (NO in step S55), the support apparatus 200 performs the most A data set of appropriate model parameters is searched for (step S57), and the searched model parameters and results of the corresponding performance evaluation are displayed (step S58).

Finally, the support apparatus 200 determines the model parameters for generating the predictive model 140 according to an instruction from the user (step S59).

In this way, the support device 200, as a function related to the generation of the predictive model 140, sequentially varies the model parameters defining the predictive model 140 under the conditions of the determined explanatory variable and the determined section, so that each model By evaluating a plurality of indices with respect to the predictive model 140 defined by the parameters, a process for determining the model parameters for defining the predictive model 140 is executed.

Fig. 21 is a diagram showing an example of a user interface provided in a process for determining a model parameter in the prediction system 1 according to the embodiment. In the user interface screen 270 shown in FIG. 21 , the prediction accuracy 281 , the model size 282 , and the processing speed 283 are calculated as compared to the user interface screen 270 shown in FIGS. 16 to 19 . ) is additionally indicated.

In this way, the support device 200 provides a user interface screen 270 (refer to FIG. 21 ) displaying a plurality of indicators. The user may adjust explanatory variables and intervals while checking these indicators.

In addition, the search for the optimal model parameter in step S57 of Fig. 20 may be realized by setting priorities for indices such as prediction accuracy, model size, and processing speed.

For example, it is assumed that the priority of the index is set in the order of prediction precision, model size, and processing speed. In addition, you may set the minimum condition about each index|index.

For example, you may make the following settings.

·Priority 1 indicator: Prediction precision condition; Condition: Prediction precision≥80%

·Priority 2 indicator: model size; Condition: Model size≤50MB

·Priority 3 indicators: processing speed; Condition: processing speed≤0.1msec

When each of these indexes is set, the support apparatus 200 extracts a set of model parameters that are good for the priority 1 index, and then searches for a data set in which both the priority 2 index and the priority 3 index are good.

In addition, a certain priority may be set with respect to a certain index|index, and you may set the same priority with respect to a some index|index. Moreover, you may determine the optimal data set by combining the results of evaluation with respect to each index|index without setting a priority.

In this way, in the determination of the model parameters (step S5), the model parameters with a good overall index (typically, high precision, small model size, and high processing speed) are determined by grid search (full search). can

In addition, the user can generate an appropriate predictive model 140 by fine-tuning model parameters, explanatory variables, and intervals as necessary, while checking the performance of the predictive model according to the determined model parameters on the user interface screen. .

By providing such a user interface screen, the prediction model 140 can be determined by comprehensive evaluation including a plurality of indices such as prediction accuracy, model size, and processing speed.

<I. Variations>

In the above description, although the prediction system 1 which predicts time change was demonstrated, it is applicable also to the abnormality detection system which detects the abnormality which arises in a control object etc.

22 is a schematic diagram showing the outline configuration of an abnormality detection system 1A according to a modification of the present embodiment. Referring to Fig. 22, the abnormality detection system 1A selects a training sample from the raw data 40 acquired from the control target (sample selection 42), and based on the selected training sample, the abnormality detection model (44) is generated. Then, using the generated abnormality detection model 44, the operation 46 of abnormality detection is executed.

The abnormality detection model 44 has the subject of detecting that the control target exhibits a state different from the normal state, and uses raw data (time-series data) collected from the control target to estimate the collected raw data. A suitable anomaly detection model 44 is created. By inputting raw data different from normal to the abnormality detection model 44, and outputting a value indicating that the state is different from normal, it can be detected that some kind of abnormality has occurred in the control target.

As for the training sample used for such an abnormality detection model 44, it is preferable to employ|adopt raw data from which a change pattern differs similarly to step S3 mentioned above.

FIG. 23 is a diagram for explaining a sample selection processing example shown in FIG. 22 . Referring to FIG. 23 , for example, when five pieces of raw data 401 to 405 are collected, the raw data 401 and the raw data 404 have a similar distribution. Similarly, the raw data 402 and the raw data 405 have a similar distribution.

As for raw data having such a mutually similar distribution, it is preferable to use only one raw data as a learning sample. As a result, in the example shown in FIG. 23 , any one of the raw data 401 and the raw data 404 , any one of the raw data 402 and the raw data 405 , and three of the raw data 403 . You may select a kind of raw data as a learning sample.

In this way, also about the training sample used for the abnormality detection model 44, similarly to step S3 mentioned above, the detection performance of the abnormality detection model 44 can be improved by extracting the sample from which a change pattern differs from each other.

<J. Bookkeeping>

The present embodiment as described above includes the following technical ideas.

[Configuration 1]

a control operation unit 164 that executes a control operation for controlling a control object;

a predicted value acquisition unit 156 for acquiring a predicted value by inputting an performance value composed of one or more state values among the state values that the control calculation unit can refer to into a prediction model;

and a predictive model generator 200 that determines the predictive model in advance,

The predictive model generation unit,

means 2261 for determining one or a plurality of state values as explanatory variables, based on the importance to the training sample, among a plurality of state values associated with a learning sample used for generating the predictive model;

Means 2261 for determining a section to be used for prediction by sequentially varying sections included in the search section and evaluating the prediction precision by the determined explanatory variable;

Under the conditions of the determined explanatory variable and the determined interval, sequentially different model parameters defining the predictive model, and evaluating a plurality of indices with respect to the predictive model defined by each model parameter, thereby defining the predictive model means (2262) for determining a model parameter for a prediction system.

[Configuration 2]

According to configuration 1, the predictive model generating unit further includes means (2261) for selecting a sample to be used as the training sample from among a plurality of samples by clustering using one or a plurality of feature quantities calculated from each sample. prediction system.

[Configuration 3]

The prediction system according to configuration 1 or 2, wherein the predictive model generation unit calculates the importance of each state value by applying a decision tree algorithm to the learning sample and a plurality of related state values.

[Configuration 4]

The prediction system according to any one of Configurations 1 to 3, wherein the predictive model generating unit evaluates the performance of the predictive model generated from each pattern in which at least one of a length and a position of a section included in the search section is different.

[Configuration 5]

The prediction system according to configuration 4, wherein the predictive model generation unit searches a section used for the prediction after fixing the model parameters of the predictive model.

[Configuration 6]

The prediction system according to configuration 4 or 5, wherein the prediction model generation unit further includes means for providing a first user interface (270) for accepting a user operation for changing a section used for prediction.

[Configuration 7]

The prediction system according to any one of Configurations 1 to 6, wherein the plurality of indexes include at least one of prediction accuracy, model size, and processing speed.

[Configuration 8]

The prediction system according to configuration 7, wherein the prediction model generating unit further includes means for providing a second user interface (270) for displaying the plurality of indices.

[Configuration 9]

An information processing device (200) connected to a control device (100), the control device comprising: a control calculation unit (164) that executes a control calculation for controlling a control target; and a predicted value acquisition unit 156 for acquiring a predicted value by inputting the performance value composed of a plurality of state values into the prediction model;

The information processing device is a predictive model generating unit that determines the predictive model in advance,

means 2261 for determining one or a plurality of state values as explanatory variables, based on the importance to the training sample, among a plurality of state values associated with a learning sample used for generating the predictive model;

Means 2261 for determining a section to be used for prediction by sequentially varying sections included in the search section and evaluating the prediction precision by the determined explanatory variable;

Under the conditions of the determined explanatory variable and the determined interval, sequentially different model parameters defining the predictive model, and evaluating a plurality of indices with respect to the predictive model defined by each model parameter, thereby defining the predictive model means (2262) for determining model parameters for

[Configuration 10]

An information processing program (226) executed on a computer (200) connected to a control device (100), the control device comprising: a control calculation unit (164) that executes a control calculation for controlling a control target; a predicted value acquisition unit 156 for acquiring a predicted value by inputting an performance value composed of one or a plurality of state values among the state values that can be referenced into the prediction model;

The information processing program, as processing for pre-determining the predictive model, to the computer,

A step (S43) of determining one or a plurality of state values as explanatory variables, based on the importance to the learning sample, among a plurality of state values associated with the learning sample used for generating the predictive model;

A step (S44) of determining a section to be used for prediction by sequentially varying sections included in the search section and evaluating the prediction precision by the determined explanatory variable;

Under the conditions of the determined explanatory variable and the determined interval, sequentially different model parameters defining the predictive model, and evaluating a plurality of indices with respect to the predictive model defined by each model parameter, thereby defining the predictive model An information processing program for executing steps (S5; S51 to S59) for determining model parameters for

<K. Advantages>

In the prediction system according to the present embodiment, the prediction model can be finally determined based on a plurality of indexes, so that it is possible to easily generate a prediction model suitable for actual operation.

It should be considered that embodiment disclosed this time is an illustration and is not restrictive in every point. The scope of the present invention is defined by the claims rather than the above description, and it is intended that all changes within the meaning and scope equivalent to the claims are included.

1: Prediction system
1A: Abnormality detection system
2, 4: Fieldbus
6: Upper Network
10: field device group
12: Remote I/O device
14: relay group
16, 124: I/O unit
18: image sensor
20: camera
22: servo driver
24: servo motor
30: press machine
31: work
32: intermediate product
33: base
34: support
35: press-fitting plate
36: drive shaft
37: motor
40, 401, 402, 403, 404, 405: raw data
42: sample selection
44: abnormal detection model
46: operation
100: control device
102, 202: processor
104: chip set
106, 206: main memory
108, 208: secondary memory
110, 214: upper network controller
112, 212: USB controller
114: memory card interface
116: memory card
118, 120: fieldbus controller
122: internal bus controller
132: memory area
134: export module
140: predictive model
142: model parameters
144: reference orbit
150: PLC engine
154: user program
156: prediction value acquisition code
158: error evaluation code
160: additional learning code
162: record code
164: control opcode
200: support device
204: optical drive
205: recording medium
216: input unit
218: display unit
220: bus
222: development program
224: interface program
226: analysis program
230: user interface
232: script engine
234: config file
236: I/O management module
238: display module
240: graph library
242: analysis module
244: analysis library
250, 260, 270: User interface screen
252, 262: composite waveform
254: sample group
264: similarity
271: prediction point
272: section
273: section width
274: time difference
275: prediction error
276: waveform display area
277: list of explanatory variables
281: Prediction precision
282: model size
283: processing speed
300: parent server
400: display device
2261: section determination module
2262: model creation module
2263: inference module
2264: evaluation module

Claims (10)

a control operation unit that executes a control operation for controlling a control object;
a predicted value acquisition unit that acquires a predicted value by inputting an performance value comprising one or a plurality of state values among the state values that the control calculating unit can refer to into a prediction model;
and a predictive model generator that determines the predictive model in advance,
The predictive model generation unit,
means for determining one or a plurality of state values as explanatory variables from among a plurality of state values associated with a learning sample used for generating the predictive model, based on the importance to the learning sample;
means for determining a section to be used for prediction by sequentially varying sections included in the search section and evaluating the prediction precision by the determined explanatory variable;
Under the conditions of the determined explanatory variable and the determined interval, sequentially different model parameters defining the predictive model, and evaluating a plurality of indices with respect to the predictive model defined by each model parameter, thereby defining the predictive model A prediction system comprising means for determining model parameters for According to claim 1,
The predictive model generating unit further includes means for selecting a sample to be used as the training sample from among a plurality of samples by clustering using one or a plurality of feature quantities calculated from each sample. According to claim 1,
The predictive model generating unit is configured to calculate the importance of each state value by applying a decision tree algorithm to the learning sample and a plurality of related state values. 4. The method according to any one of claims 1 to 3,
The prediction model generation unit, the prediction system, characterized in that each evaluates the performance of the prediction model generated in each pattern in which at least one of the length and position of the section included in the search section is different. 5. The method of claim 4,
The prediction system generating unit, characterized in that after fixing the model parameter of the prediction model, searching for a section used for the prediction. 5. The method of claim 4,
The prediction system according to claim 1, wherein the prediction model generation unit further includes means for providing a first user interface for accepting a user operation for changing a section used for prediction. According to any one of claims 1 to 3,
The prediction system, characterized in that the plurality of indices include at least one of prediction accuracy, model size, and processing speed. 8. The method of claim 7,
The prediction system according to claim 1, wherein the predictive model generating unit further includes means for providing a second user interface for displaying the plurality of indices. An information processing device connected to a control device, the control device comprising: a control calculation unit that executes a control operation for controlling a control object; A predicted value acquisition unit for acquiring a predicted value by inputting to
The information processing device is a predictive model generator that determines the predictive model in advance,
means for determining one or a plurality of state values as explanatory variables from among a plurality of state values associated with a learning sample used for generating the predictive model, based on the importance to the learning sample;
means for determining a section to be used for prediction by sequentially varying sections included in the search section and evaluating the prediction precision by the determined explanatory variable;
Under the conditions of the determined explanatory variable and the determined interval, sequentially different model parameters defining the predictive model, and evaluating a plurality of indices with respect to the predictive model defined by each model parameter, thereby defining the predictive model An information processing device comprising means for determining a model parameter for A recording medium storing an information processing program executed by a computer connected to a control device, the control device comprising: a control arithmetic unit that executes a control arithmetic for controlling a control object; A predicted value acquisition unit for acquiring a predicted value by inputting the performance value composed of a plurality of state values into the prediction model;
The information processing program, as processing for determining the predictive model in advance, to the computer,
determining one or a plurality of state values as explanatory variables based on the importance to the learning sample among a plurality of state values associated with a learning sample used for generating the predictive model;
determining a section to be used for prediction by sequentially varying sections included in the search section and evaluating the prediction precision by the determined explanatory variable;
Under the conditions of the determined explanatory variable and the determined interval, sequentially different model parameters defining the predictive model, and evaluating a plurality of indices with respect to the predictive model defined by each model parameter, thereby defining the predictive model A recording medium for executing a step of determining model parameters for

Images